We met with two business leaders in the last few days and got talking about repetitive work in their respective businesses. This is common, but both conversations landed somewhere new for us:

“It's a design we've sold [a client] 24 times in the past, and we're somehow still taking 3 or 4 days to get it out. Where can we automate?”

- CTO, Design Manufacturer

“The fact that we collect, you know, 10,000 pictures of retail shelves a day allows us a lot of information to be able to go back to the brands and help them. That’s what’s useful.”

- CEO & Founder, CPG Specialist

Last week’s post was about the work nobody wants to do (Diplomas Collecting Dust), and while these quotes reference repetitive work there’s another step. Both the designer and the CPG company should absolutely automate those processes and move on. That’s what they’re now doing, but automating is becoming the easy part. That makes the process on its own a weaker “edge” than before.

What’s hard to copy is the byproduct from all that repetitive work. Sure, anyone with a Claude Code subscription and a free weekend can hack together some sort of automation, but they don’t have 1,000,000+ photos of store shelves like the CPG guy quoted or enough designs to need multiple storage systems, which is what the manufacturer is dealing with. AI can’t replicate that.

Repetitive tasks often require manual input and lead to some form of data. Most teams are so focused on getting rid of the manual input that they forget about the data, which is actually valuable.

So, sure, start with the small wins and automate away certain functions. That said, build each automation knowing what you’re truly protecting because we’ve seen two ways that this goes wrong:

1. Over-simplifying: in the spirit of “streamlining” the repetitive task shrinks and so does the result. Certain data tags get dropped, filing habits get ignored, or the overall volume decreases.

2. Lowering standards: getting 90% right sounds pretty good for a basic automation, but that’s a risk when the 10% of errors pollutes the output quality. Once the 10% of bad eggs get mixed in with the rest because it was “good enough”, the resulting data stops being a trusted source.

So here's the next step to take after reading this:

• Select the next workflow or process that you’re planning to automate. Instead of focusing on simplifying the process, start by writing out every attribute of the finished data when the process is done manually. Next to each attribute (e.g., tag, notes, codes, approvals) put a check mark if it’s still captured by the automated version.

We’d suggest only building automations that have checkmarks for every feature. Yes, the mind-numbing work needs to go. No, the data being collected shouldn’t deteriorate in the process.